# OpenAI Pulls Cursor’s Models: 95% of Users Feel Nothing > Source: > Published: 2026-09-02 15:11:15+00:00 OpenAI announced on August 28 that it is terminating Cursor’s access to all its AI models by November 12, 2026 — invoking a change-of-control clause after SpaceX’s $60 billion acquisition of Anysphere, Cursor’s parent company, closed on August 14. The stated reason: OpenAI [cannot be confident](https://openai.com/index/our-decision-on-cursor-following-its-acquisition-by-spacex/) that SpaceX will use its technology within its terms of service, citing Elon Musk’s companies’ history of violating contracts. For millions of developers who use Cursor daily, the news sounds alarming. The actual disruption is much narrower — but the precedent being set is worth paying attention to. ## OpenAI’s Case Is About Musk’s History, Not Cursor’s Conduct Cursor itself hasn’t violated anything. OpenAI’s termination is guilt by ownership. The company is invoking prior violations by other Musk-controlled entities: under oath earlier this year, Musk admitted that xAI “distilled” OpenAI’s model outputs to help train Grok — a direct breach of OpenAI’s terms of service, which prohibit using outputs to train competing models. Twitter/X separately cut off OpenAI’s data access after the acquisition while allegedly scraping OpenAI API outputs. OpenAI’s position is that with xAI now part of SpaceX, the same ownership structure that produced those violations now owns Cursor. It’s a legally defensible argument. Whether it’s a wise one is a different question. [According to CNBC](https://www.cnbc.com/2026/08/29/openai-cursor-spacex-model-access.html), Cursor CEO Michael Truell has said his team is in talks with OpenAI to resolve the situation. Don’t hold your breath — OpenAI’s future models, including Astra, are also off the table for Cursor under the current position. ## The 5% Number Changes Everything Here’s the stat that puts this story in perspective: OpenAI models account for roughly [5% of Cursor user traffic](https://x.com/mntruell/status/2093532254006063557), according to Truell himself. The remaining 95% runs on Claude (Anthropic) or Gemini (Google), neither of which is going anywhere. If you opened Cursor today and your default model was Claude Sonnet — which describes the majority of active users — November 12 changes nothing about your workflow. The developers who actually need to act are those who explicitly configured GPT-4o, o3, or a direct OpenAI enterprise API key as their Cursor model. For them, it’s a settings change: switch to Claude, Gemini, or the new Grok Bot that SpaceXAI launched as its first post-acquisition product. That’s not nothing, especially if you’ve built specific prompt workflows around GPT-4o behavior. However, it’s not the infrastructure collapse the headlines suggest. Related:[Cursor /multitask: Parallel Agents and the Cost You’re Not Tracking] ## Anthropic’s Two-Hour Response Was Not Accidental Less than two hours after OpenAI published its statement, Anthropic co-founder Tom Brown posted that Cursor has been a trusted partner since Sonnet 3.5 and pledged to increase compute to support Claude models in Cursor. That’s not a coincidence of timing — it’s a prepared play. Anthropic knew OpenAI’s move was coming and had a response ready to deploy. The context makes Brown’s confidence understandable. Anthropic’s Q2 2026 revenue reached $11.5 billion, surpassing OpenAI’s $6.7 billion for the first time. Claude already powers the vast majority of Cursor traffic. OpenAI just handed Anthropic a cleaner lock on one of the world’s most-used developer tools — and Anthropic is expanding compute to reinforce it. In the AI coding market, this is what winning looks like without winning a fight. ## The Real Risk: AI Model Providers as Political Actors This is the part of the story that deserves more attention than the Musk drama. OpenAI’s decision establishes a new norm: AI model providers can terminate access to developer tools based on who owns those tools, not how the tools themselves use the models. That’s a fundamentally different risk profile than a product deprecation or a price change. As [Forkast notes](https://forkast.news/openais-cursor-termination-turns-model-supply-into-a-competitive-weapon/), this is model supply being turned into a competitive weapon. For enterprises evaluating AI coding tools, this adds a new due-diligence question: what happens to this tool’s model access if it gets acquired by someone in a legal or competitive dispute with a major AI provider? Multi-model support — the ability to switch between Claude, Gemini, Grok, and others — is no longer just a convenience feature. It’s a resilience requirement. Cursor already supports this. The tools that don’t are the ones worth worrying about. ## Key Takeaways - OpenAI is cutting Cursor’s model access by November 12, 2026, citing prior ToS violations by Musk-owned xAI and Twitter/X — not anything Cursor itself did. - Only 5% of Cursor traffic runs on OpenAI models. If you’re using Claude or Gemini as your default, nothing changes on November 12. - Developers using GPT-4o or direct OpenAI API keys in Cursor need to migrate their model configuration before the cutoff date. - Anthropic responded within hours with expanded Claude compute in Cursor and already leads OpenAI in revenue ($11.5B vs. $6.7B in Q2 2026). - The bigger issue: AI model access is now a political variable. Multi-model support in coding tools isn’t optional anymore — it’s supply-chain resilience.